Abstract
East Asians are known to develop diabetes mellitus at a lower body weight than Caucasians, potentially because of the different mechanisms underlying disease development. This study aimed to evaluate the variation in weight transition leading to diabetes onset in two subtypes of individuals (obese and non-obese) in a Japanese population. We conducted a retrospective, observational, longitudinal cohort study using health checkup data from 9, 260 participants in Japan. Individuals who developed diabetes within three years of the start of the observation period were excluded. Among the participants, 61.4% were men, and 259 developed diabetes. In the obesity group (body mass index [BMI] ≥25 kg/m2), the average BMI increased prior to the diabetes onset and subsequently decreased. Conversely, in the non-obesity group (BMI <25 kg/m2), the average BMI decreased and then stabilized before the onset of diabetes. Notably, a greater number of participants in the non-obesity group exhibited a BMI change of ≤–0.15 kg/m2 per year compared with those with a BMI change of ≥0.15 kg/m2 per year before diabetes onset (p = 0.003). Our findings indicate that body weight loss precedes the onset of diabetes in the non-obesity group. We recommend that non-obese individuals with elevated blood glucose levels who do not meet the criteria for diabetes should be considered a high-risk group for diabetes development. Therefore, it is imperative to identify these individuals and provide lifestyle guidance that does not focus on weight loss to prevent the onset of diabetes.
Keywords: Body mass index, Type 2 diabetes mellitus
Graphical Abstract
Introduction
Diabetes mellitus affects approximately 537,000,000 individuals globally [1]. This number is expected to increase to 643,000,000 by 2030. East Asians, including the Japanese population, exhibit a higher susceptibility to type 2 diabetes than Caucasians [2], whereas type 1 diabetes is more prevalent among Caucasians than among East Asians. In Japan, the prevalence of diabetes was reported to be 11,000,000 in 2021 [1], with dietary westernization potentially contributing to this increase [3]. Since 2008, Japan has implemented Specific Health Checkups and Specific Health Guidance for middle-aged individuals to facilitate the early detection and prevention of metabolic syndrome [4]. Given the association between metabolic syndrome and diabetes [5], these health initiatives aim to prevent diabetes by promoting weight loss.
Traditionally, diabetes has been classified into type 1, type 2, and other specific types (e.g., monogenic diabetes syndromes, diseases of the exocrine pancreas, and drug- or chemical-induced diabetes) [6]. Type 1 diabetes is characterized by autoimmune destruction of pancreatic beta cells, leading to insulin deficiency, while type 2 diabetes is primarily due to a combination of insulin resistance and insufficient insulin secretion. However, there has recently been a shift toward a more nuanced classification based on pathogenesis [7], such as increased insulin resistance and decreased insulin secretory capacity [8]. When the classification proposed by Ahlqvist et al. [8] was applied to the Japanese population, the incidence of diabetic complications differed between groups [9].
We hypothesized that a significant proportion of Japanese individuals with diabetes exhibit impaired insulin secretion independent of obesity. Previous studies have shown that Asians generally have a lower insulin secretory capacity than Caucasians [10]. Additionally, the risk of developing diabetes increases at a body mass index (BMI) of ≥25 kg/m2 in Caucasians, whereas in Asians, this risk threshold is lower at a BMI of ≥23 kg/m2 [11, 12]. Therefore, the current Specific Health Guidance may be inadequate for preventing diabetes in non-obese individuals.
This study focused on two subtypes of diabetes: one characterized by increased insulin resistance and the other by decreased insulin secretion. We analyzed the factors contributing to diabetes development in each group using BMI as a surrogate marker for the diabetes subtype, owing to the challenges of directly measuring insulin secretion in large populations. Given the positive correlation between BMI and insulin resistance [13], we expected that obese individuals would develop diabetes primarily because of increased insulin resistance. In contrast, non-obese individuals would be more likely to develop diabetes because of decreased insulin secretion.
This study aimed to investigate the patterns of weight change among obese and non-obese cohorts before the diagnosis of diabetes mellitus and to explore the implications of personalized medicine in diabetes prevention.
Materials and Methods
Study design and ethical approval
This retrospective, observational, longitudinal cohort study used data from annual and biannual health checkups provided by the Hokuriku Health Service Association in Toyama Prefecture, Japan. All procedures adhered to the 1964 Helsinki Declaration and its subsequent amendments, as well as the “Ethical Guidelines for Medical and Biological Research Involving Human Subjects” issued by the Ministry of Health, Labour, and Welfare of Japan. The study protocol was approved by the Ethics Committee of Toyama University Hospital (approval number: R2021070). Written informed consent was obtained from all participants for the use of their data in this study.
Study participants
The participants were individuals who underwent health checkups at the Hokuriku Health Service Association, which serves approximately 130,000–170,000 individuals annually. From April 2012 to March 2013, 138,378 individuals (men: 89,705 [64.8%], mean age: 42.7 ± 13.0 years) who underwent health checkups were analyzed. Among them, we initially selected 18,373 participants aged 49–64 years between April 2012 and March 2021. Although the cohort was not restricted to the Japanese population, the number of non-Japanese participants was expected to be low. Participants who could not be followed up for more than three years due to reasons such as job changes or relocation were excluded. Additionally, patients who developed diabetes within three years of the start of the observation period were excluded. Finally, 12,635 participants were included in the study, and fasting data were available for only 9,260 participants. The year in which each participant developed diabetes was designated as the year of onset (year 0), as illustrated in the individual participant timeline (Supplementary Fig. 1). The analysis was conducted in two parts, as presented in the overview of the first and second analyses (Supplementary Fig. 2). The first analysis included 9,260 participants with fasting data, and the second analysis included all 12,635 participants.
Data collection
Health checkup data, including self-reported sex, age, physical data, laboratory data, and questionnaire responses regarding lifestyle habits, were collected and analyzed for the fiscal years 2012–2020. The physical data included the BMI, abdominal circumference (AC), systolic blood pressure (sBP), and diastolic blood pressure (dBP). Laboratory data included total cholesterol (T-Cho), high-density lipoprotein cholesterol (HDL-Cho), low-density lipoprotein cholesterol (LDL-Cho), triglyceride (Tg), plasma glucose (PG), glycosylated hemoglobin (HbA1c), uric acid, creatinine (Cre), aspartate aminotransferase (AST), alanine aminotransferase (ALT), and γ-glutamyl transpeptidase (γ-GTP). The questionnaire data included smoking status; use of antihypertensive, diabetes, and dyslipidemia medications; history of stroke and cardiac disease; weight gain of more than 10 kg since the age of 20 years; engagement in light sweating exercise (at least 30 minutes per session, at least twice a week, for at least one year); eating speed compared to others; eating dinner within two hours before bedtime at least three times a week; frequency of alcohol consumption; and motivation to improve lifestyle. A total of 53,195 fasting and 42,656 non-fasting blood samples were collected and analyzed.
Definition of terms
Diabetes was diagnosed based on health checkup data, if participants met any of the following criteria according to the Japanese Clinical Practice Guideline for Diabetes [14]: (a) fasting PG level ≥126 mg/dL and HbA1c ≥6.5%, (b) non-fasting PG level ≥200 mg/dL and HbA1c ≥6.5%, (c) self-reported history of diabetes, or (d) self-reported diabetes treatment. The first year in which a participant was diagnosed with diabetes was considered the onset year. Participants diagnosed with diabetes were assumed to have diabetes without recovery from the disease. The obesity group was defined as participants with an average BMI of ≥25 kg/m2 during the observation period, while the non-obesity group had an average BMI of <25 kg/m2. Previous studies reported a BMI increase of 0.20 (95% confidence interval [CI]: 0.15–0.24) per year in individuals developing diabetes [15]. Our study observed significant variability in BMI change, with a median of 0.2 and a mean of 0.06. Therefore, the weight gain group (BWG) was defined as participants with an annual increase of ≥0.15 kg/m2. The body weight loss group (BWL) comprised participants with an annual BMI decrease of >0.15 kg/m2. The body weight stable group (BWSt) comprised participants who did not meet the criteria for BWG or BWL. The hyperglycemia group included participants with high blood glucose levels (fasting PG level ≥110 mg/dL or non-fasting PG level ≥140 mg/dL) at the beginning of the observation period who did not meet the diagnostic criteria for diabetes.
Statistical analysis
Student’s t-test and the chi-square test were used to compare numerical and categorical variables, respectively. Binomial tests were used to assess the equality of frequencies for binary variables. Time series data were analyzed with the diabetes onset year set at year 0. Spearman’s correlation coefficient was used to analyze the correlations between years and HbA1c, years and BMI, and years and AC from the start of observation to onset of diabetes, excluding the post-onset period. The coefficients were calculated using a 95% CI (2.5%–97.5%). Numerical variables were presented as mean ± standard deviation. Missing values were excluded from each analysis, and missing BMIs were imputed using the previous year’s BMI. Statistical significance was set at p < 0.05. All statistical analyses were performed using Python version 3.8.8 (Python Software Foundation, Beaverton, OR, USA) [16] and TableOne (Python Software Foundation, Beaverton, OR, USA) [17].
Results
Clinical characteristics of the study participants and comparison of the diabetes group and non-diabetes group
Table 1 presents the clinical characteristics of the study participants with available fasting data (first analysis). Among the participants (n = 9,260), 61.4% were men, with a mean age of 55.9 ± 3.2 years, a mean BMI of 22.9 ± 3.4 kg/m2, a mean AC of 83.4 ± 9.3 cm, a mean PG level of 94.6 ± 12.8 mg/dL, and a mean HbA1c level of 39 ± 4 mmol/mol (5.7 ± 0.4%). The basis for the diagnosis of diabetes, including duplicates, was abnormal laboratory values (152 participants, 59%) and questionnaire responses (diabetes or medication history) (115 participants, 44%). Compared with the non-diabetes group (n = 9,001), the diabetes group (n = 259, 2.80%) exhibited a higher percentage of men (80.3% vs. 60.9%, p < 0.001), higher BMI (25.9 ± 3.9 vs. 22.8 ± 3.3 kg/m2, p < 0.001), higher AC (91.0 ± 9.9 vs. 83.1 ± 9.2 cm, p < 0.001), higher PG levels (122.6 ± 27.3 vs. 93.6 ± 10.8 mg/dL, p < 0.001), and higher HbA1c levels (48 ± 8 mmol/mol (6.5 ± 0.8%) vs. 38 ± 3 mmol/mol (5.6 ± 0.3%), p < 0.001). A similar trend was observed when the obesity group (n = 2,185) and the non-obesity group (n = 7,075) were separately analyzed (Supplementary Table 1). The incidence of diabetes was higher in the obesity group compared with the non-obesity group (6.73% vs. 1.58%, p < 0.001). Among non-obese individuals, 17% in the hyperglycemia group developed diabetes during the observation period, whereas the incidence of diabetes among non-obese individuals outside the hyperglycemia group was 1.2%. The mean observation period for each case was 8.13 ± 1.48 years, with each case examined approximately 11 times, resulting in a total of 102,777 examinations.
Table 1. Clinical characteristics of the study participants and comparison of the diabetes group and the non-diabetes group during the entire observation period.
| All (n = 9,260) |
Diabetes (n = 259) |
Non-diabetes (n = 9,001) |
Missing n (%) |
p-value (diabetes vs. non-diabetes) |
|
|---|---|---|---|---|---|
| Male (n) | 5,693 (61.4 %) | 208 (80.3 %) | 5,485 (60.9 %) | 0 (0%) | <0.001 |
| Age (years) | 55.9 ± 3.2 | 56.3 ± 3.2 | 55.9 ± 3.2 | 0 (0%) | <0.001 |
| Smoking† (n) | 15,017 (27.9 %) | 615 (34.8 %) | 14,402 (27.7 %) | 73 (0.14%) | <0.001 |
| BMI (kg/m2) | 22.9 ± 3.4 | 25.9 ± 3.9 | 22.8 ± 3.3 | 0 (0%) | <0.001 |
| AC (cm) | 83.4 ± 9.3 | 91.0 ± 9.9 | 83.1 ± 9.2 | 208 (0.39%) | <0.001 |
| sBP (mmHg) | 124.9 ± 16.0 | 130.1 ± 15.2 | 124.7 ± 16.0 | 31 (0.057%) | <0.001 |
| dBP (mmHg) | 77.3 ± 11.5 | 81.1 ± 10.2 | 77.2 ± 11.6 | 31 (0.057%) | <0.001 |
| T-Cho (mg/dL) | 215.2 ± 34.1 | 212.8 ± 37.6 | 215.3 ± 34.0 | 3,119 (5.8%) | 0.008 |
| HDL-Cho (mg/dL) | 63.9 ± 16.5 | 53.6 ±14.0 | 64.3 ± 16.5 | 60 (0.11%) | <0.001 |
| LDL-Cho (mg/dL) | 130.3 ± 31.7 | 132.1 ± 33.8 | 130.2 ± 31.6 | 60 (0.11%) | 0.022 |
| Tg (mg/dL) | 116.1 ± 97 | 169.2 ± 146.8 | 114.3 ± 94.3 | 60 (0.11%) | <0.001 |
| PG (mg/dL) | 94.6 ± 12.8 | 122.6 ± 27.3 | 93.6 ± 10.8 | 1,103 (2.0%) | <0.001 |
| HbA1c (mmol/mol) | 39 ± 4 | 48 ± 8 | 38 ± 3 | 17,127 (32%) | <0.001 |
| HbA1c (%) | 5.7 ± 0.4 | 6.5 ± 0.8 | 5.6 ± 0.3 | 17,127 (32%) | <0.001 |
| UA (mg/dL) | 5.6 ± 1.4 | 5.9 ± 1.4 | 5.6 ± 1.4 | 4,634 (8.6%) | <0.001 |
| Cre (mg/dL) | 0.8 ± 0.3 | 0.8 ± 0.2 | 0.8 ± 0.3 | 5,503 (10%) | 0.002 |
| AST (U/L) | 24.2 ± 15.8 | 30.0 ± 39.4 | 24.0 ± 14.3 | 56 (0.10%) | <0.001 |
| ALT (U/L) | 23.5 ± 19.2 | 36.8 ± 26.5 | 23.1 ± 18.7 | 56 (0.10%) | <0.001 |
| γ-GTP (U/L) | 45.6 ± 59.3 | 72.2 ± 83.3 | 44.7 ± 58.1 | 56 (0.10%) | <0.001 |
The statistics of each variable were calculated using the data from the entire observation period. BMI, body mass index; AC, abdominal circumference; sBP, systolic blood pressure; dBP, diastolic blood pressure; T-Cho, total-cholesterol; HDL-Cho, high density lipoprotein-cholesterol; LDL-Cho, low density lipoprotein-cholesterol; Tg, triglyceride; PG, plasma glucose; HbA1c, glycosylated hemoglobin; UA, uric acid; Cre, creatinine; AST, aspartate aminotransferase; ALT, alanine aminotransferase; γ-GTP, γ-glutamyl transpeptidase. † Total number of participants who responded that they smoke cigarettes.
Comparison of the obesity group and non-obesity group in the diabetes group
We compared the obesity group and non-obesity group in the diabetes group (Table 2). Compared with the non-obesity group (n = 112), the obesity group (n = 147) exhibited higher BMI (28.5 ± 2.8 vs. 22.4 ± 1.9 kg/m2, p < 0.001), higher AC (96.9 ± 7.8 vs. 83.1 ± 6.1 cm, p < 0.001), higher sBP (131.2 ± 14.5 vs. 128.7 ± 16.0 mmHg, p = 0.001), higher ALT level (39.8 ± 24.4 vs. 32.9 ± 28.7 U/L, p < 0.001), and lower γ-GTP level (65.2 ± 53.5 vs. 81.5 ± 110.3 U/L, p < 0.001). Additionally, the obesity group had a higher proportion of participants taking antihypertensive medication (41.9% vs. 19.9%, p < 0.001), dyslipidemia medication (22.1% vs. 10.7%, p < 0.001), and >10 kg weight gain since age 20 (81.7% vs. 36.2%, p < 0.001). Interestingly, the obesity group had a lower proportion of smokers (29.9% vs. 41.3%, p < 0.001), a higher proportion of participants engaging in light sweating exercise habits (12.1% vs. 8.9%, p = 0.051), a lower proportion of daily alcohol consumption (32.8% vs. 44.6%, p < 0.001), and a lower proportion of participants with no intention to improve their lifestyle (22.9% vs. 41.9%, p < 0.001).
Table 2. Comparison of the obesity group and the non-obesity group in the diabetes group during the entire observation period.
| Obesity (n = 147) |
Non-obesity (n = 112) |
Missing n (%) |
p-value | |
|---|---|---|---|---|
| Percentage of diabetes incidence (%) | 6.73 % | 1.58 % | <0.001 | |
| Male (n) | 117 (79.6 %) | 91 (81.3 %) | 0 (0%) | 0.593 |
| Age (years) | 56.2 ± 3.3 | 56.4 ± 3.2 | 0 (0%) | 0.125 |
| Smoking† (n) | 300 (29.9 %) | 315 (41.3 %) | 3 (0.17%) | <0.001 |
| BMI (kg/m2) | 28.5 ± 2.8 | 22.4 ± 1.9 | 0 (0%) | <0.001 |
| AC (cm) | 96.9 ± 7.8 | 83.1 ± 6.1 | 5 (0.28%) | <0.001 |
| sBP (mmHg) | 131.2 ± 14.5 | 128.7 ± 16.0 | 3 (0.17%) | 0.001 |
| dBP (mmHg) | 82.0 ± 10.0 | 80.0 ± 10.4 | 3 (0.17%) | <0.001 |
| T-Cho (mg/dL) | 213.8 ± 37.1 | 211.5 ± 38.1 | 66 (3.73%) | 0.221 |
| HDL-Cho (mg/dL) | 50.8 ± 11.7 | 57.3 ± 15.9 | 2 (0.11%) | <0.001 |
| LDL-Cho (mg/dL) | 134.1 ± 33.9 | 129.4 ± 33.6 | 2 (0.11%) | 0.004 |
| Tg (mg/dL) | 184.6 ± 164.8 | 148.8 ± 116.1 | 2 (0.11%) | <0.001 |
| PG (mg/dL) | 123.3 ± 29.1 | 121.7 ± 24.8 | 30 (1.70%) | 0.201 |
| HbA1c (mmol/mol) | 48 ± 8 | 48 ± 9 | 619 (35%) | 0.67 |
| HbA1c (%) | 6.5 ± 0.8 | 6.5 ± 0.9 | 619 (35%) | 0.67 |
| UA (mg/dL) | 6.1 ± 1.4 | 5.7 ± 1.5 | 150 (8.48%) | <0.001 |
| Cre (mg/dL) | 0.8 ± 0.2 | 0.8 ± 0.1 | 174 (9.84%) | <0.001 |
| AST (U/L) | 30.1 ± 15.2 | 29.9 ± 57.4 | 1 (0.057%) | 0.916 |
| ALT (U/L) | 39.8 ± 24.4 | 32.9 ± 28.7 | 1 (0.057%) | <0.001 |
| γ-GTP (U/L) | 65.2 ± 53.5 | 81.5 ± 110.3 | 1 (0.057%) | <0.001 |
| Antihypertensive medication (n) | 385 (41.9 %) | 137 (19.9 %) | 161 (9.11%) | <0.001 |
| Diabetes medications (n) | 152 (16.6 %) | 119 (17.3 %) | 161 (9.11%) | 0.756 |
| Dyslipidemia medications (n) | 203 (22.1 %) | 74 (10.7 %) | 161 (9.11%) | <0.001 |
| History of stroke (n) | 14 (1.5 %) | 12 (1.7 %) | 161 (9.11%) | 0.888 |
| History of cardiac disease (n) | 59 (6.4 %) | 24 (3.5 %) | 161 (9.11%) | 0.012 |
| Weight gain of more than 10 kg since age 20 (n) | 748 (81.7 %) | 247 (36.2 %) | 170 (9.62%) | <0.001 |
| Light sweating exercise for at least 30 minutes at a time, at least 2 days a week, for at least 1 year (n) | 111 (12.1 %) | 61 (8.9 %) | 171 (0.67%) | 0.051 |
| Eating speed compared to others (n) | ||||
| Fast | 501 (55.1 %) | 189 (27.8 %) | 177 (10.01%) | <0.001 |
| Normal | 368 (40.4 %) | 440 (64.6 %) | ||
| Slow | 41 (4.5 %) | 52 (7.6 %) | ||
| Eating dinner within 2 hours before bedtime at least 3 times a week (n) | 260 (28.4 %) | 156 (22.9 %) | 170 (9.62%) | 0.015 |
| Frequency of alcohol consumption (n) | ||||
| Every day | 301 (32.8 %) | 307 (44.6 %) | 162 (9.16%) | <0.001 |
| Occasionally | 161 (17.6 %) | 97 (14.1 %) | ||
| Rare | 455 (49.6 %) | 285 (41.4 %) | ||
| Motivation to improve lifestyle (n) | ||||
| No intention of improving | 208 (22.9 %) | 285 (41.9 %) | 177 (10.01%) | <0.001 |
| Intending to improve (within 6 months) | 346 (38.0 %) | 199 (29.2 %) | ||
| Intend to improve soon (within 1 month) | 104 (11.4 %) | 54 (7.9 %) | ||
| Already working on improving (within 6 months) | 100 (11.0 %) | 42 (6.2 %) | ||
| Already working on improving (over 6 months) | 152 (16.7 %) | 101 (14.8 %) | ||
The statistics of each variable were calculated using the data from the entire observation period. BMI, body mass index; AC, abdominal circumference; sBP, systolic blood pressure; dBP, diastolic blood pressure; T-Cho, total-cholesterol; HDL-Cho, high density lipoprotein-cholesterol; LDL-Cho, low density lipoprotein-cholesterol; Tg, triglyceride; PG, plasma glucose; HbA1c, glycosylated hemoglobin; UA, uric acid; Cre, creatinine; AST, aspartate aminotransferase; ALT, alanine aminotransferase; γ-GTP, γ-glutamyl transpeptidase. † Total number of participants who responded that they smoke cigarettes.
Comparison of the obesity group and the non-obesity group regarding parameter changes before and after the diabetes onset
We conducted a comparative analysis of the obesity and non-obesity groups with respect to changes in HbA1c levels, BMI, and AC before and after the onset of diabetes (Fig. 1). Given that HbA1c, BMI, and AC values were independent of the timing of blood sampling, we performed the analysis without considering the timing of blood sampling (second analysis). The analysis included 389 participants who developed diabetes and were divided into the obesity group (n = 215) and non-obesity group (n = 174). In the obesity group, HbA1c levels, BMI, and AC increased before the onset of diabetes, whereas BMI and AC decreased after the onset. Conversely, in the non-obesity group, HbA1c levels increased before the onset of diabetes, whereas BMI and AC decreased and subsequently leveled off. When only fasting blood samples were analyzed, the results were similar for the obesity group, and there was no significant increase or decrease in BMI and AC before the onset of diabetes in the non-obesity group (Supplementary Fig. 3). Similar results were obtained when participants were divided into three groups (Supplementary Fig. 4). In the group with an average BMI of less than 23 kg/m2, BMI and AC decreased and then leveled off before the onset of diabetes. Due to missing HbA1c data in 32% of the records, additional analyses were conducted, in which diabetes was redefined based solely on the use of diabetes medications and a history of diabetes. Similarly, in the non-obesity group, BMI (r = –0.10, p = 0.008) and AC (r = –0.10, p = 0.01) decreased and then leveled off before the onset of diabetes (Supplementary Fig. 5). We also examined the longitudinal changes in people without diabetes. In contrast to those who developed diabetes, both the obesity and non-obesity groups exhibited increases in HbA1c levels, BMI, and waist circumference over time. The average BMI of non-obese individuals without diabetes decreased slightly from the first to the fourth year of observation. However, the change was much smaller than the early BMI change in non-obese individuals who developed diabetes (Supplementary Fig. 6).
Fig. 1. Changes in average parameters before and after the diabetes onset in the obesity group (a) and the non-obesity group (b). Dashed lines represent the year of diabetes onset, designated as year 0. Data were analyzed using Spearman’s rank correlation coefficients. BMI, body mass index; AC, abdominal circumference; HbA1c, glycosylated hemoglobin.
Detailed comparison of the obesity group and the non-obesity group regarding weight changes before the diabetes onset
We further examined the weight change before the onset of diabetes in the obesity and non-obesity groups using a binomial test (Fig. 2). The participants were classified into three groups based on weight change before the onset of diabetes: BWG, BWSt, and BWL. In the non-obesity group (n = 174), a higher proportion of participants were classified as having BWL than as having BWG (38% vs. 20%, p = 0.003). Additionally, since the average BMI in the non-obesity group largely decreased 7–8 years before the onset of diabetes (Fig. 1), we separately analyzed the participants observed for 7–8 years before the onset (n = 61), and those observed 3–6 years before the onset of diabetes (n = 113). In both cases, the binomial test indicated that a higher proportion of participants were classified as BWL than as BWG (7–8 year observation: 31% vs. 12%, p = 0.029; 3–6 year observation: 41% vs. 25%, p = 0.037). Interestingly, in the obesity group (n = 215), the proportions of BWL and BWG were comparable (28% vs. 31%, p = 0.659). In summary, BWL was prevalent in the non-obesity group, but it was also observed in the obesity group.
Fig. 2. Comparison of weight changes before diabetes onset between the obesity and non-obesity groups. P-values were calculated based on the null hypothesis that the frequencies of the BWG and BWL participants were equal. BMI, body mass index; BWG, body weight gain; BWSt, body weight stable; BWL, body weight loss.
Discussion
In this study, we analyzed the health checkup data of 9,260 participants in Japan (first analysis) and 12,635 participants in Japan (second analysis). The first analysis confirmed that the diabetes group exhibited a higher percentage of men, elevated BMI, and increased AC than the non-diabetes group (Table 1). This corroborates the established understanding that diabetes is more prevalent in men and is associated with obesity [18, 19]. Additionally, the diabetes-obesity group demonstrated higher sBP, a greater proportion of individuals taking antihypertensive medication, and a higher proportion of individuals taking dyslipidemia medication than the diabetes-non-obesity group (Table 2). This supports the notion that obesity, in conjunction with diabetes, is associated with hypertension and dyslipidemia. Furthermore, the diabetes-obesity group was significantly more likely to have experienced weight gain of >10 kg since the age of 20 years than the diabetes-non-obesity group (Table 2), suggesting that both environmental and genetic factors contribute to the long-term differences between these groups. Weight gain since youth has been identified as a risk factor for developing diabetes [20]. Interestingly, the diabetes-non-obesity group exhibited a higher proportion of smokers, higher proportion of daily alcohol consumers, lower proportion of individuals engaging in light sweating exercise, and a higher proportion of individuals with no intention of improving their lifestyle than the diabetes-obesity group (Table 2). These findings imply that the non-obese participants who developed diabetes during the observation period had unhealthy lifestyle habits. Smoking is a risk factor for type 2 diabetes [21]. Although the relationship between alcohol consumption and the risk of type 2 diabetes follows a J-shaped curve [22], moderate to high alcohol consumption has been associated with type 2 diabetes in non-obese middle-aged Japanese individuals [23].
In the second analysis, we observed that BMI and AC changed in opposite directions in the obesity and non-obesity groups before the onset of diabetes (Fig. 1). Specifically, BMI and AC increased before the onset of diabetes in the obesity group but decreased before the onset of diabetes in the non-obesity group. The former phenomenon aligns with the general belief that diabetes primarily develops when insulin secretion from pancreatic β-cells cannot compensate for obesity-induced insulin resistance. Although seemingly paradoxical, the latter phenomenon suggests that in the non-obesity group analyzed in this study, diabetes was primarily caused by a decrease in insulin secretion rather than insulin resistance. Because increased insulin secretion generally leads to body weight gain through the stimulation of glucose uptake in tissues and the promotion of lipogenesis, particularly in adipocytes, it is plausible that a decrease in insulin secretion may result in weight loss. Body weight at the time of diabetes diagnosis is known to reflect insulin secretory capacity. Even in obese individuals, BMI and AC appear to decrease just before diabetes onset. However, given that health checkups are annual, it is assumed that in some cases, diabetes is diagnosed and treatment is initiated, whereas in others, glucose metabolism deteriorates, leading to weight loss.
Therefore, we hypothesized that in a state of excess energy, body weight increases when insulin resistance is adequately compensated by insulin secretion from pancreatic β-cells. Conversely, when insulin secretion cannot be compensated, catabolism occurs, and body weight gradually decreases. Thus, we speculate that the extent of weight gain or loss before developing diabetes may depend on the endogenous insulin secretory capacity. Additionally, if weight loss serves as a marker of decreased insulin secretion, this study suggests that insulin secretion in the non-obesity group begins to decline at least eight years before diabetes onset. This hypothesis is supported by a previous study that reported a decrease in insulin secretion ten years before the development of diabetes [24].
Many non-obese participants developed diabetes during our study because our data targeted the Japanese population only. A previous study suggested that being underweight may be a risk factor for diabetes in Asians [25]. Another study reported that underweight young Japanese women exhibited impaired glucose tolerance, characterized by decreased insulin secretory capacity, increased insulin resistance, and reduced energy intake and activity [26]. We hypothesized that East Asians, including Japanese, are more likely to develop diabetes with a low BMI because of their lower endogenous insulin secretory capacity than that of Caucasians. A previous study reported that Caucasians had a higher pancreatic β-cell mass than East Asians and that the response of pancreatic β-cells to obesity differed between Caucasians and Asians [10]. These differences may contribute to racial disparities in body weight before diabetes development.
The different scenarios leading to the development of diabetes in obese and non-obese individuals (Fig. 1) and Caucasians and Asians may have a genetic basis. Previous studies have identified gene regions associated with diabetes development [27-29]. These studies suggest that insulin secretory capacity, tissues involved in insulin sensitivity, such as muscle and fat, and microRNAs contribute to the development of type 2 diabetes. Racial differences in gene regions related to diabetes development have also been observed. In the future, we aim to identify the genetic factors that explain the differences in diabetes pathogenesis and calculate the corresponding polygenic risk scores [30].
Therefore, non-obese individuals with high PG levels who do not meet the criteria for diabetes mellitus should be considered as a high-risk group for developing diabetes. Furthermore, it is essential to recognize these individuals and provide appropriate lifestyle guidance to prevent the development of diabetes. However, in Japan, Specific Health Checkups and Specific Health Guidance primarily focus on metabolic syndrome [4], and the risk of developing diabetes in non-obese individuals has been overlooked. Non-obese individuals were unaware of the need to change their lifestyles, indicating that they had unhealthy habits (Table 2). Based on existing diabetes prevention programs [31], including multidisciplinary interventions in diet and exercise, it is necessary to identify and improve each factor involved in the development of future diabetes in individual cases, rather than solely aiming for weight loss.
This study had some limitations. First, data were collected from a single health checkup association, the Hokuriku Health Service Association, which primarily targets workers and their families in Toyama Prefecture, Japan. Consequently, selection bias must be considered when interpreting these results. Second, the type of diabetes was not specified, preventing the differentiation between type 2 and type 1 diabetes. However, given the lower prevalence of type 1 diabetes in Japan than in Caucasian populations [32], it is anticipated that most diabetes cases in this study were of type 2 diabetes. Third, there were instances of missing data, as some participants did not undergo regular health checkups. Additionally, certain health check items, including HbA1c, were optional and not consistently available at all health checkup sites. Fourth, detailed information on alcohol consumption and physical activity could not be obtained because of the self-administered nature of the questionnaire. Fifth, direct measurement of insulin secretion was not feasible. Sixth, data on race were not collected. Seventh, several risk factors for the development of diabetes were not evaluated. Specifically, we could not assess fatty liver [33], visceral fat accumulation [34], or family history of diabetes mellitus [35], which have been identified as risk factors for diabetes development in previous studies.
In the future, we propose conducting a prospective study to investigate the risk factors for the development of diabetes, including measurements of insulin secretion, in the high-risk groups identified in this study.
In conclusion, the analysis of Japanese health checkup data revealed distinct weight trends related to the onset of diabetes. In the non-obesity group, weight decreased and then stabilized before the onset of diabetes, whereas in the obesity group, weight increased before diabetes onset (Graphical Abstract). Specific Health Guidance was implemented for individuals with metabolic syndrome and non-obese individuals were generally excluded. We recommend that non-obese individuals with elevated PG levels who do not meet the criteria for diabetes should be considered as a high-risk group for diabetes development. Given that the current forms of Specific Health Guidance are insufficient to prevent the development of diabetes in the high-risk group identified in this study, it is imperative to provide new lifestyle guidance that does not focus solely on weight loss.
Graphical Abstract.
Acknowledgements
The authors thank the clinical staff who contributed to the health checkups.
We would like to thank Editage (www.editage.jp) for English language editing.
Data availability
The data that support the findings of this study are not openly available for reasons of sensitivity, but are available from the corresponding author upon reasonable request. Data were stored in controlled-access data storage at the Laboratory of Preventive Medicine, Hokuriku Health Service Association.
Disclosure
Compliance with ethical standards
Approval of the research protocol: The Ethics Committee of Toyama University Hospital approved the study protocol (approval number: R2021070; date of approval: 8/19/2021).
Informed Consent: The study participants consented to the use of their data for scientific research.
Approval date of Registry and the Registration No. of the study/trial: N/A
Animal Studies: N/A
Disclosure of potential conflicts of interest
Kazuyuki Tobe has received lecture fees from Sumitomo Pharma Co. Ltd. and Novo Nordisk Pharma Ltd.; grants from Tokai National Higher Education and Research System and the University of Tokyo; and scholarship donations from Eli Lilly Japan K.K., the Japan Association for Diabetes Education and Care, and the Uehara Memorial Foundation. The authors declare no potential conflict of interest.
Funding sources
This study was supported by the JST Moonshot Research and Development Program (Grant Number JPMJMS2021).The study funder was not involved in the design of the study, collection, analysis, and interpretation of data, writing of the report, and did not impose any restrictions regarding the publication of the report.
Supplementary Material
Supplementary Table
Comparison of the diabetes group and the non-diabetes group in the obesity group and the non-obesity group
Supplementary Figs.
References
- 1.International Diabetes Federation (2021) IDF Diabetes Atlas (10th). International Diabetes Federation, Brussels, Belgium. [Google Scholar]
- 2.Kodama K, Tojjar D, Yamada S, Toda K, Patel CJ, et al. (2013) Ethnic differences in the relationship between insulin sensitivity and insulin response: a systematic review and meta-analysis. Diabetes Care 36: 1789–1796. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Yabe D, Seino Y, Fukushima M, Seino S (2015) β cell dysfunction versus insulin resistance in the pathogenesis of type 2 diabetes in East Asians. Curr Diab Rep 15: 602. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Tsushita K, Hosler AS, Miura K, Ito Y, Fukuda T, et al. (2018) Rationale and descriptive analysis of specific health guidance: the nationwide lifestyle intervention program targeting metabolic syndrome in Japan. J Atheroscler Thromb 25: 308–322. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Alberti KG, Eckel RH, Grundy SM, Zimmet PZ, Cleeman JI, et al. (2009) Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation 120: 1640–1645. [DOI] [PubMed] [Google Scholar]
- 6.American Diabetes Association Professional Practice Committee (2022) 2. Classification and diagnosis of diabetes: standards of medical care in diabetes—2022. Diabetes Care 45: S17–S38. [DOI] [PubMed] [Google Scholar]
- 7.Chung WK, Erion K, Florez JC, Hattersley AT, Hivert MF, et al. (2020) Precision medicine in diabetes: a consensus report from the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). Diabetes Care 43: 1617–1635. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Ahlqvist E, Storm P, Käräjämäki A, Martinell M, Dorkhan M, et al. (2018) Novel subgroups of adult-onset diabetes and their association with outcomes: a data-driven cluster analysis of six variables. Lancet Diabetes Endocrinol 6: 361–369. [DOI] [PubMed] [Google Scholar]
- 9.Tanabe H, Saito H, Kudo A, Machii N, Hirai H, et al. (2020) Factors associated with risk of diabetic complications in novel cluster-based diabetes subgroups: a Japanese retrospective cohort study. J Clin Med 9: 2083. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Inaishi J, Saisho Y (2020) Beta-cell mass in obesity and type 2 diabetes, and its relation to pancreas fat: a mini-review. Nutrients 12: 3846. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hsu WC, Araneta MR, Kanaya AM, Chiang JL, Fujimoto W (2015) BMI cut points to identify at-risk Asian Americans for type 2 diabetes screening. Diabetes Care 38: 150–158. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Araneta MR, Kanaya AM, Hsu WC, Chang HK, Grandinetti A, et al. (2015) Optimum BMI cut points to screen asian americans for type 2 diabetes. Diabetes Care 38: 814–820. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Farin HM, Abbasi F, Reaven GM (2006) Body mass index and waist circumference both contribute to differences in insulin-mediated glucose disposal in nondiabetic adults. Am J Clin Nutr 83: 47–51. [DOI] [PubMed] [Google Scholar]
- 14.Araki E, Goto A, Kondo T, Noda M, Noto H, et al. (2020) Japanese Clinical Practice Guideline for Diabetes 2019. Diabetol Int 11: 165–223. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Hu H, Kawasaki Y, Kuwahara K, Nakagawa T, Honda T, et al. (2020) Trajectories of body mass index and waist circumference before the onset of diabetes among people with prediabetes. Clin Nutr 39: 2881–2888. [DOI] [PubMed] [Google Scholar]
- 16.Van Rossum G, Drake FL (2009) Python 3 reference manual. CreateSpace, Scotts Valley, USA. [Google Scholar]
- 17.Pollard TJ, Johnson AEW, Raffa JD, Mark RG (2018) tableone: an open source Python package for producing summary statistics for research papers. JAMIA Open 1: 26–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Nanri A, Nakagawa T, Kuwahara K, Yamamoto S, Honda T, et al. (2015) Development of risk score for predicting 3-year incidence of type 2 diabetes: Japan epidemiology collaboration on occupational health study. PLoS One 10: e0142779. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Hu H, Nakagawa T, Yamamoto S, Honda T, Okazaki H, et al. (2018) Development and validation of risk models to predict the 7-year risk of type 2 diabetes: The Japan epidemiology collaboration on occupational health study. J Diabetes Investig 9: 1052–1059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Sun W, Shi L, Ye Z, Mu Y, Liu C, et al. (2016) Association between the change in body mass index from early adulthood to midlife and subsequent type 2 diabetes mellitus. Obesity (Silver Spring) 24: 703–709. [DOI] [PubMed] [Google Scholar]
- 21.Pan A, Wang Y, Talaei M, Hu FB, Wu T (2015) Relation of active, passive, and quitting smoking with incident type 2 diabetes: a systematic review and meta-analysis. Lancet Diabetes Endocrinol 3: 958–967. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Han M (2020) The dose-response relationship between alcohol consumption and the risk of type 2 diabetes among Asian men: a systematic review and meta-analysis of prospective cohort studies. J Diabetes Res 2020: 1032049. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Waki K, Noda M, Sasaki S, Matsumura Y, Takahashi Y, et al. (2005) Alcohol consumption and other risk factors for self-reported diabetes among middle-aged Japanese: a population-based prospective study in the JPHC study cohort I. Diabet Med 22: 323–331. [DOI] [PubMed] [Google Scholar]
- 24.Ohn JH, Kwak SH, Cho YM, Lim S, Jang HC, et al. (2016) 10-year trajectory of β-cell function and insulin sensitivity in the development of type 2 diabetes: a community-based prospective cohort study. Lancet Diabetes Endocrinol 4: 27–34. [DOI] [PubMed] [Google Scholar]
- 25.Jung JY, Park SK, Oh CM, Ryoo JH, Choi JM, et al. (2018) The risk of type 2 diabetes mellitus according to the categories of body mass index: the Korean Genome and Epidemiology Study (KoGES). Acta Diabetol 55: 479–484. [DOI] [PubMed] [Google Scholar]
- 26.Sato M, Tamura Y, Nakagata T, Someya Y, Kaga H, et al. (2021) Prevalence and features of impaired glucose tolerance in young underweight Japanese women. J Clin Endocrinol Metab 106: e2053–e2062. [DOI] [PubMed] [Google Scholar]
- 27.Yamauchi T, Hara K, Maeda S, Yasuda K, Takahashi A, et al. (2010) A genome-wide association study in the Japanese population identifies susceptibility loci for type 2 diabetes at UBE2E2 and C2CD4A-C2CD4B. Nat Genet 42: 864–868. [DOI] [PubMed] [Google Scholar]
- 28.Suzuki K, Akiyama M, Ishigaki K, Kanai M, Hosoe J, et al. (2019) Identification of 28 new susceptibility loci for type 2 diabetes in the Japanese population. Nat Genet 51: 379–386. [DOI] [PubMed] [Google Scholar]
- 29.Spracklen CN, Horikoshi M, Kim YJ, Lin K, Bragg F, et al. (2020) Identification of type 2 diabetes loci in 433,540 East Asian individuals. Nature 582: 240–245. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Khera AV, Chaffin M, Aragam KG, Haas ME, Roselli C, et al. (2018) Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations. Nat Genet 50: 1219–1224. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Lindström J, Peltonen M, Eriksson JG, Ilanne-Parikka P, Aunola S, et al. (2013) Improved lifestyle and decreased diabetes risk over 13 years: long-term follow-up of the randomised Finnish Diabetes Prevention Study (DPS). Diabetologia 56: 284–293. [DOI] [PubMed] [Google Scholar]
- 32.Forouhi NG, Wareham NJ (2014) Epidemiology of diabetes. Medicine (Abingdon) 42: 698–702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Narisada A, Shibata E, Hasegawa T, Masamura N, Taneda C, et al. (2021) Sex differences in the association between fatty liver and type 2 diabetes incidence in non-obese Japanese: A retrospective cohort study. J Diabetes Investig 12: 1480–1489. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Tang Y, Wei ZM, Li N, Sun LL, Jin ZY, et al. (2022) Quantitative analysis of the risk of type 2 diabetes and fatty liver in non-obese individuals by computed tomography. Abdom Radiol (NY) 47: 2099–2105. [DOI] [PubMed] [Google Scholar]
- 35.Iwata M, Kamura Y, Honoki H, Kobayashi K, Ishiki M, et al. (2020) Family history of diabetes in both parents is strongly associated with impaired residual β-cell function in Japanese type 2 diabetes patients. J Diabetes Investig 11: 564–572. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Table
Comparison of the diabetes group and the non-diabetes group in the obesity group and the non-obesity group
Supplementary Figs.
Data Availability Statement
The data that support the findings of this study are not openly available for reasons of sensitivity, but are available from the corresponding author upon reasonable request. Data were stored in controlled-access data storage at the Laboratory of Preventive Medicine, Hokuriku Health Service Association.